Bias and ratios: reading a number without being fooled

Entraînement. Une question à la fois, correction immédiate. Enchaînez autant de questions que vous voulez. Rien n'est enregistré, pas besoin de compte.

Scoring. Loan approval: Men 52% (n=2,300), Women 44% (n=2,100), "not reported" 9% (n=41). The report shows an overall disparity ratio of 5.8. Fairest course of action?

Group	n	Rate
Men	2,300	52%
Women	2,100	44%
not reported	41	9%

Use 5.8 as the primary indicator and demand immediate mitigation Exclude "not reported" from the indicator and examine the M/F gap (~1.18) as the true disparity Weight each group by its headcount to smooth out the small group's effect, then re-conclude Conclude there is no bias: 44/52 = 0.85 ≥ 0.80, the four-fifths rule is satisfied